1 |
>>>>> On Mon, 18 May 2009, Ciaran McCreesh wrote: |
2 |
|
3 |
>> Trouble starts if hyphens in PV are allowed. |
4 |
|
5 |
> You mean like -r0? |
6 |
|
7 |
The revision is not part of PV. And it's easily split off, since the |
8 |
string "-r<digits>" cannot occur elsewhere in the package version. |
9 |
|
10 |
> It's easily solved by a careful definition, in any case, just the |
11 |
> same way that there's already a careful definition full of |
12 |
> weaselling out to allow other abuses... There's no ambiguity so long |
13 |
> as the definition is sound. |
14 |
|
15 |
To come back to my example: |
16 |
|
17 |
>> P=foo-1a-scm could mean both of the following: |
18 |
>> |
19 |
>> PN=foo PV=1a-scm |
20 |
>> PN=foo-1a PV=scm |
21 |
|
22 |
AFAICS, there _is_ an ambiguity. You can have the following two |
23 |
ebuilds in the tree, simultaneously: |
24 |
|
25 |
${PORTDIR}/app-misc/foo/foo-1a-scm.ebuild |
26 |
${PORTDIR}/app-misc/foo-1a/foo-1a-scm.ebuild |
27 |
|
28 |
Which package will be pulled in by the following dependency? |
29 |
|
30 |
RDEPEND="=app-misc/foo-1a-scm" |
31 |
|
32 |
The conclusion is that GLEP 54 in its current form is not implementable. |
33 |
(But maybe it would be possible to use a period instead of the hyphen? |
34 |
That is, ".live" instead of "-scm"?) |
35 |
|
36 |
Ulrich |